In this thesis, I present two methods: the Relative Energy Gradient (REG) Method and the FFLUX machine learning force-field. The Relative Energy Gradient (REG) Method answers two questions: Which subsets of partitioned energies best describe the total behaviour of a chemical system? and Can chemical insight be gained from subsets of partitioned energies? It is shown in this thesis that both of these questions are answered by the REG Method when used in conjunction with Interacting Quantum Atoms (IQA) approach. By using the IQA method, a system (such as a molecule) can be partitioned into subsets of atomistic energies that recover the total energy when summed. Applying the REG method to the partitioned IQA energies allows for the easy, automated analysis of a system. As such, arbitrarily sized systems can be studied in an exhaustive manner. The FFLUX force-field is a next generation force field currently under development. The methodology used to calculate energies and forces in FFLUX differs from traditional force fields, in that it does not require the use of harmonic potentials and instead uses the machine learning method Kriging to predict IQA energies. Because of this, the FFLUX force field does not require any empirical parameterisation and is able to perform calculations at near- quantum accuracy. During my PhD I have implemented FFLUX in the molecular dynamics program DL_POLY and used it to geometry optimise a peptide-capped glycine molecule, the results of which are given in this thesis.